diff --git a/Lib/test/test_fstring.py b/Lib/test/test_fstring.py index b41e02c3a16379..41cefe0e286d50 100644 --- a/Lib/test/test_fstring.py +++ b/Lib/test/test_fstring.py @@ -1831,6 +1831,34 @@ def test_newlines_in_format_specifiers(self): for case in valid_cases: compile(case, "", "exec") + def test_raw_fstring_format_spec(self): + # Test raw f-string format spec behavior (Issue #137314). + # + # Raw f-strings should preserve literal backslashes in format specifications, + # not interpret them as escape sequences. + class UnchangedFormat: + """Test helper that returns the format spec unchanged.""" + def __format__(self, format): + return format + + # Test basic escape sequences + self.assertEqual(f"{UnchangedFormat():\xFF}", 'ÿ') + self.assertEqual(rf"{UnchangedFormat():\xFF}", '\\xFF') + + # Test nested expressions with raw/non-raw combinations + self.assertEqual(rf"{UnchangedFormat():{'\xFF'}}", 'ÿ') + self.assertEqual(f"{UnchangedFormat():{r'\xFF'}}", '\\xFF') + self.assertEqual(rf"{UnchangedFormat():{r'\xFF'}}", '\\xFF') + + # Test continuation character in format specs + self.assertEqual(f"""{UnchangedFormat():{'a'\ + 'b'}}""", 'ab') + self.assertEqual(rf"""{UnchangedFormat():{'a'\ + 'b'}}""", 'ab') + + # Test multiple format specs in same raw f-string + self.assertEqual(rf"{UnchangedFormat():\xFF} {UnchangedFormat():\n}", '\\xFF \\n') + if __name__ == '__main__': unittest.main() diff --git a/Misc/NEWS.d/next/Core_and_Builtins/2025-08-02-23-04-57.gh-issue-137314.wjEdzD.rst b/Misc/NEWS.d/next/Core_and_Builtins/2025-08-02-23-04-57.gh-issue-137314.wjEdzD.rst new file mode 100644 index 00000000000000..09d0c3e68fc1ed --- /dev/null +++ b/Misc/NEWS.d/next/Core_and_Builtins/2025-08-02-23-04-57.gh-issue-137314.wjEdzD.rst @@ -0,0 +1,5 @@ +Fixed a regression where raw f-strings incorrectly interpreted +escape sequences in format specifications. Raw f-strings now properly preserve +literal backslashes in format specs, matching the behavior from Python 3.11. +For example, ``rf"{obj:\xFF}"`` now correctly produces ``'\\xFF'`` instead of +``'ÿ'``. Patch by Pablo Galindo. diff --git a/Parser/action_helpers.c b/Parser/action_helpers.c index ebc94715b6f361..0763866576733f 100644 --- a/Parser/action_helpers.c +++ b/Parser/action_helpers.c @@ -1404,7 +1404,15 @@ expr_ty _PyPegen_decoded_constant_from_token(Parser* p, Token* tok) { if (PyBytes_AsStringAndSize(tok->bytes, &bstr, &bsize) == -1) { return NULL; } - PyObject* str = _PyPegen_decode_string(p, 0, bstr, bsize, tok); + + // Check if we're inside a raw f-string for format spec decoding + int is_raw = 0; + if (INSIDE_FSTRING(p->tok)) { + tokenizer_mode *mode = TOK_GET_MODE(p->tok); + is_raw = mode->raw; + } + + PyObject* str = _PyPegen_decode_string(p, is_raw, bstr, bsize, tok); if (str == NULL) { return NULL; }